Using MCP to Add Visual Proof to Your Cursor Automations

Cursor just launched Automations: write a task description, and Cursor's agent goes — navigating, clicking, filling forms, extracting data.

It's powerful. But how do you know the agent actually did what you asked?

That's where MCP (Model Context Protocol) comes in. MCP lets you extend Cursor's capabilities with custom tools. And PageBolt is an MCP server that adds visual proof to every automation.

What Is MCP?

MCP is a protocol that lets Claude (and other AI clients) talk to external tools and services. Instead of building everything into Cursor itself, you can write an MCP server and plug it in.

Think of it like REST APIs, but for Claude.

With MCP, you can:

• Connect Cursor to your internal tools
• Add custom integrations to Cursor Automations
• Extend what Cursor can do without waiting for Cursor to build it

PageBolt's MCP server is one example: it gives Cursor the ability to take screenshots and record videos.

Installing PageBolt's MCP Server

Add PageBolt to your Cursor config:

{
  "mcp": {
    "servers": {
      "pagebolt": {
        "command": "npx",
        "args": ["@pagebolt/mcp-server"],
        "env": {
          "PAGEBOLT_API_KEY": "pf_live_YOUR_API_KEY"
        }
      }
    }
  }
}

Restart Cursor. Now your automations have access to PageBolt's tools.

Adding Visual Proof to Automations

Write an automation that uses PageBolt:

Task: Fill out the contact form on example.com

Steps:
1. Navigate to https://example.com/contact
2. Take a screenshot using PageBolt to show the form
3. Fill in "Name" with "John Doe"
4. Fill in "Email" with "john@example.com"
5. Fill in "Message" with "Hello, I'm interested in your services"
6. Take a screenshot to show the filled form
7. Click "Submit"
8. Take a screenshot to show the success page
9. Record the entire session as video using PageBolt

Output: Return the screenshot URLs and video URL as proof

When Cursor runs this automation, it:

• Executes the steps using its browser tools
• Calls PageBolt's screenshot tool after each major step
• Records the entire session as video
• Returns the proof artifacts in the output

You now have visual evidence of what the automation did.

Why This Matters

Cursor Automations are powerful, but they're opaque. You get a result, but no visibility into what happened.

Visual proof solves this:

Debugging: When an automation fails, watch the video and see exactly where it went wrong.

Governance: When your team asks "what did that automation actually do?", show them the screenshots and video.

Compliance: When regulators ask for proof of execution, you have pixel-perfect evidence.

Trust: Stakeholders trust automations they can see working.

Real-World Example

Automation: Process customer support tickets from a web portal

Without visual proof:

• Ticket ID: TICKET-001
• Status: Closed
• Result: Success

With visual proof:

• Ticket ID: TICKET-001
• Screenshot 1: Form before fill
• Screenshot 2: Form after fill
• Screenshot 3: Submission confirmation page
• Video: Full 2-minute session recording

When your support team audits the process, they can watch exactly what happened.

Getting Started

1. Install PageBolt's MCP server in your Cursor config
2. Get a free API key at pagebolt.dev
3. Write an automation that includes screenshot/video steps
4. Run the automation and collect visual proof
5. Store the proofs with your automation results for auditing

Cursor gives you browser automation. MCP extends it. PageBolt adds visibility.

Together, they're the full stack for trustworthy automation.

---